Graphic processing units (GPUs) are rapidly gaining maturity as powerful general par-allel computing devices. A key feature in the development of modern GPUs has been theadvancement of the programming model and programming tools. Compute Uniﬁed De-vice Architecture (CUDA) is a software platform for massively parallel high-performancecomputing on Nvidia many-core GPUs. In functional magnetic resonance imaging (fMRI),the volume of the data to be processed, and the type of statistical analysis to perform callfor high-performance computing strategies. In this work, we present the main features ofthe R-CUDA package cudaBayesreg which implements in CUDA the core of a Bayesianmultilevel model for the analysis of brain fMRI data. The statistical model implementsa Gibbs sampler for multilevel/hierarchical linear models with a normal prior. The maincontribution for the increased performance comes from the use of separate threads forﬁtting the linear regression model at each voxel in parallel. The R-CUDA implementationof the Bayesian model proposed here has been able to reduce signiﬁcantly the run-timeprocessing of Markov chain Monte Carlo (MCMC) simulations used in Bayesian fMRIdata analyses. Presently, cudaBayesreg is only conﬁgured for Linux systems with NvidiaCUDA support.
|Journal||Journal Of Statistical Software|
|Publication status||Published - 1 Jan 2011|